NVIDIA and SB Energy Partner with OpenAI for Massive AI Factory in Ohio
- The initial deployment at the Ohio campus is expected to provide 4.25 gigawatts of AI factory capacity, backed by NVIDIA support for infrastructure payments and residual-value commitments totaling...
- The PORTS-Pike Technology Campus will utilize NVIDIA's full-stack DSX AI factory platform, incorporating advanced graphics processing units, central processing units, networking gear, and infrastructure software, according to NVIDIA.
- Lease, land, and power capacity has emerged as a critical bottleneck for artificial intelligence development, particularly for frontier labs whose infrastructure demands outpace their balance sheets.
The initial deployment at the Ohio campus is expected to provide 4.25 gigawatts of AI factory capacity, backed by NVIDIA support for infrastructure payments and residual-value commitments totaling billions of dollars through 2030.
PORTS-Pike Technology Campus Details and Capacity
The PORTS-Pike Technology Campus will utilize NVIDIA’s full-stack DSX AI factory platform, incorporating advanced graphics processing units, central processing units, networking gear, and infrastructure software, according to NVIDIA. Each generation of hardware deployed at the Portsmouth site could support roughly 1.5 million NVIDIA graphics processing units per cycle. According to the company, these deployments represent approximately $150 billion to $200 billion in revenue per upgrade cycle, with the campus designed to support multiple upgrade cycles over a 20-year term. In addition to the initial 4.25 gigawatts, NVIDIA retains the option to extend the Portsmouth arrangement to secure an additional 3.75 gigawatts of remaining capacity at the site. OpenAI has committed to substantial deployments of NVIDIA infrastructure extending through 2030, representing roughly 12 gigawatts of total compute that could expand to 16 gigawatts if the PORTS-Pike agreement is fully extended. These combined commitments represent an estimated $600 billion business opportunity for NVIDIA through the end of the decade, according to the source material.
Infrastructure Constraints and Frontier AI Lab Demand
Lease, land, and power capacity has emerged as a critical bottleneck for artificial intelligence development, particularly for frontier labs whose infrastructure demands outpace their balance sheets. While the world’s largest cloud service providers utilize independent balance sheets and long-term capital contracts to build AI factories independently, many frontier labs grow faster than their credit profiles allow. According to NVIDIA, these labs face growth constraints dictated by compute availability rather than customer demand or algorithmic limits. To address this resource gap, NVIDIA is applying supply-chain discipline to secure lease, land, and power capacity exclusively for NVIDIA-powered AI factories. The company stated that its financial support at PORTS-Pike is limited to defined portions of lease and power payments alongside a specified residual-value commitment rather than covering the full cost of the site. NVIDIA’s exposure is structured to decline over time as data centers come into service between 2028 and 2030 and OpenAI makes scheduled lease payments.
Fungibility and Ecosystem Versatility of NVIDIA Compute
Addressing questions regarding potential future vacancy at the Portsmouth campus, NVIDIA emphasized the versatile and fungible nature of its accelerated computing hardware. Because the CUDA programming platform provides a unified ecosystem for developers and engineers, deployed systems maintain broad market utility across cloud providers, enterprises, and startups. According to NVIDIA, if a primary tenant alters its footprint, the underlying infrastructure capacity can be resold and redeployed to other qualified operators across the global ecosystem without stranding the capital investment.